Skip to main content

K230-Vision

Overview

K230-Vision is RockBase IoT's AI vision module, built on Canaan's next-generation K230 dual-core RISC-V processor with up to 6 TOPS equivalent AI compute (13.7× the K210). The module runs a Linux + RT-Smart dual system and supports CanMV / MicroPython development, enabling multimodal AI applications including visual recognition, speech processing, and image understanding.

K230-Vision comes in Standard and Deluxe editions; the Deluxe edition integrates a 3.5-inch capacitive touchscreen and works out of the box. Its serial port connects directly to mainstream platforms such as ESP32, STM32, Raspberry Pi, and Jetson.

Repository: RockBase-iot/K230-Vision

Hardware Specifications

ItemSpecification
SoCCanaan K230 dual-core RISC-V
AI compute6 TOPS equivalent (13.7× K210)
Memory1 GB LPDDR4 (2 GB optional)
StorageSD card expansion
OSLinux + RT-Smart dual system
DevelopmentCanMV, MicroPython, C/C++ SDK
CameraDVP / MIPI interfaces, multi-channel input
DisplayDeluxe: 3.5" capacitive touchscreen (Standard: none)
SerialUART / USB; pairs with ESP32 / STM32 / Raspberry Pi / Jetson
NetworkEthernet (Wi-Fi 6 module / 4G expansion)
PowerUSB-C 5 V or DC 5 V input

Edition Comparison

FeatureStandardDeluxe
K230 processor
1 GB LPDDR4✓ (2 GB optional)
3.5" capacitive touchscreen
Out-of-box AI demosBasic visual recognitionMultimodal interaction (vision + speech + touch)
Best forEmbedded vision nodes, industrial inspectionHMI terminals, teaching demos, edge-AI gateways

Interfaces

Serial Connection

K230-Vision communicates with an external host over UART. Typical wiring:

K230 pinExternal host (e.g. ESP32)Description
TXRX (e.g. GPIO3)K230 data out
RXTX (e.g. GPIO1)K230 data in
GNDGNDCommon ground
3.3V3.3VLevel reference (optional)
tip

The K230-Vision serial port uses 3.3 V TTL levels and connects directly to 3.3 V devices such as ESP32, STM32, and Raspberry Pi.

Camera Interface

  • Supports DVP parallel cameras and MIPI CSI cameras
  • Recommended resolutions: 640 × 480, 1280 × 720, 1920 × 1080
  • Multi-channel camera input (Deluxe edition can run vision + touch interaction simultaneously)

Development

CanMV

CanMV is a MicroPython-based development framework for the K230 with rich AI model APIs:

# CanMV example: face detection
from machine import FPIOA
from media.sensor import SENSOR
import KPU as kpu

sensor = SENSOR()
sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)

task = kpu.load(0x300000) # load face-detection model

while True:
img = sensor.snapshot()
code = kpu.run_yolo2(task, img)
if code:
for i in code:
img.draw_rectangle(i.rect())

MicroPython

K230-Vision supports standard MicroPython syntax plus K230-specific hardware modules:

# UART example
from machine import UART

uart = UART(1, baudrate=115200, tx=2, rx=3)
uart.write(b'Hello from K230!')

AI Capabilities

CapabilityDescriptionTypical applications
Visual recognitionFace detection, object classification, pose estimationSecurity monitoring, industrial QC
Speech processingKeyword wakeup, ASR, TTSSmart speakers, voice assistants
Image understandingImage captioning, scene analysis, OCRDocument digitization, driver assistance
Edge inference6 TOPS local compute, no cloud neededOffline AI, data-privacy protection

Getting Started

1. Preparation

ItemQty
K230-Vision module×1
USB-C data cable×1
Micro-SD card (≥16 GB)×1
(Optional) camera module×1

2. Flash the System Image

  1. Download the latest system image from K230-Vision Releases
  2. Write the image to the SD card with balenaEtcher or Rufus
  3. Insert the SD card and power on

3. Connect the Serial Console

# Linux/macOS
screen /dev/ttyUSB0 115200

# Windows (PuTTY or MobaXterm)
# Select the corresponding COM port, baud rate 115200

4. Run Your First AI Demo

# Enter the CanMV environment after login
canmv

# Run the built-in face-detection demo
exec(open('/app/face_detection.py').read())

Resources

note

K230-Vision is RockBase IoT's latest AI vision platform — documentation and examples are continuously updated. Watch the GitHub repository for the latest news.